HOW THE NUMBERS WORK
EARNED ON
THE FIELD.
How strong has a team shown itself to be against the opposition it has faced?
The calculation works, but its coefficients, confidence estimates, and win probabilities have not been calibrated against historical NJ seasons. The 2025–26 archive now shows provisional Jersey Gridiron rankings calculated from published scores. The 2026–27 rankings use current published results, with automatic score updates currently paused. Early-season comparisons carry substantial uncertainty. A revised model is being tested separately in the validation preview.
The opponent matters.
Each game connects two teams. We fit the whole network together, so an earlier win can become more impressive when that opponent improves. Schedule strength is shown separately; it is not added again as a bonus.
Performance beats appearances.
A close win against a much weaker team can lower a rating. A close loss against a strong team can raise it. Group size, popularity, fan votes, and championship labels add no points.
Blowouts have limits.
The first 14 points of margin count directly. Beyond that, each extra point has less influence. Overtime is treated as a one-point adjusted margin. Home teams receive a 2.5-point adjustment; neutral sites receive none.
This season takes over.
Starting ratings use documented evidence, or an average baseline when none exists. Their direct weight declines from 70% after one game to 35% after three, 8% after six, and 1% after ten. These are starting coefficients, subject to historical testing.
Reliable scores come first.
Only verified finals from approved evidence tiers affect ratings. Disputed reports, suspicious scores, duplicates, canceled games, and unfinished games are held out. Verified forfeits affect the record, but never the performance rating.
One rating. Every leaderboard.
The statewide calculation is filtered into groups and North/South regions. It is independent of NJSIAA playoff qualification and seeding. Teams with little evidence or poorly connected schedules are marked with lower confidence.
DIMINISHING MARGINS
A bigger win.
A smaller extra reward.
Adjusted margin is a model input. It is different from the final-score margin and from the public 0–100 rating.
About the 2025–26 archive
These are retrospective Jersey Gridiron ratings calculated from MaxPreps game reports retrieved in September 2026, not MaxPreps rankings or ratings issued during the 2025 season. Identical duplicate reports count once; conflicting scores are withheld. Source confidence is an intake setting, and model confidence measures schedule evidence, not proven predictive accuracy.
Schedules include reported out-of-state games and forfeits. Those games do not contribute to the power rating without calibrated external opponents. Records in the ranking table count accepted results; profiles count individual reported games and flag disagreements with publisher summaries. School logos are the researched current marks, including labelled host-school and school-network exceptions.
What the model can—and cannot—say yet
The public rating is a smooth conversion to a 0–100 scale. Rankings use the full underlying values. Tiny differences should be read as close comparisons, not a guarantee that one team would win.
Recent results receive slightly more weight: the oldest games retain at least 80% of the recency weight. A bye has no automatic penalty. Every accepted correction recalculates the network; historical cutoffs use only revisions available at that time.
Out-of-state opponents need documented ratings calibrated to the same scale before their results enter the calculation. Quarter-by-quarter adjustments, offense and defense ratings, verified live feeds, official weekly releases, and a protected administrator dashboard are later work.
Before release, the model needs several seasons of verified data and week-by-week testing with no future results available to the calculation. Winner accuracy, margin error, and probability calibration must be measured on held-out seasons.